|
 
 |
| EDITORIAL |
|
| Year : 2021 | Volume
: 2
| Issue : 2 | Page : 25-27 |
|
Machine learning to deep learning: Artificially intelligent approaches toward precision in public health
Arista Lahiri1, Sweety Suman Jha2
1 Department of Community Medicine, College of Medicine and Sagore Dutta Hospital, Kolkata, West Bengal, India
2 Department of Preventive and Social Medicine, All India Institute of Hygiene and Public Health, Kolkata, West Bengal, India
| Date of Submission | 15-Nov-2020 |
| Date of Acceptance | 07-Dec-2020 |
| Date of Web Publication | 26-Jul-2021 |
Correspondence Address:
Dr. Arista Lahiri
37/3/1 Jaffarpur Road, 1st Lane, Barrackpore, West Bengal
India
 Source of Support: None, Conflict of Interest: None
DOI: 10.4103/jphpc.jphpc_6_20
Machine learning is in fact an application of Artificial Intelligence (AI) . It encompasses the use of algorithms in understanding the available information, i.e., data and analyzing it to arrive at an “intelligent” conclusion. Applications of AI in public health have already brought about a paradigm shift in the thinking for the provision of health care. With the global goal of universal health care, AI systems in public health can be considered very important in the resource-poor underserved areas to make a systematic arrangement for health-care delivery. The primary health care is cardinal to achieve universal health coverage. The AI systems can help the resource-contained and the grass-root level settings with remote access, algorithm-driven diagnostic aids, notification regarding emerging threats, and automated analysis of the health data in defined regions.
Keywords: Artificial intelligence, digital health, machine learning, primary care, public health
How to cite this article:
Lahiri A, Jha SS. Machine learning to deep learning: Artificially intelligent approaches toward precision in public health. J Public Health Prim Care 2021;2:25-7 |
How to cite this URL:
Lahiri A, Jha SS. Machine learning to deep learning: Artificially intelligent approaches toward precision in public health. J Public Health Prim Care [serial online] 2021 [cited 2023 May 29];2:25-7. Available from: http://www.jphpc.org/text.asp?2021/2/2/25/322307 |
| Background | |  |
Back in 1955, the term “Artificial Intelligence” (AI) was coined with unforeseen enthusiasm and optimism of changing the fabric of science and civilization.[1] Since then, the field of AI has undergone phasic expansion and retrenchment, with scientists putting their thoughts in a more intense and organized manner over the years. Following the “AI winter”[2] and the underfunded days in AI research globally, the scales tipped in favor of this enigma with the beginning of the 21st century. Interestingly, by this time, the computing power available has also improved exponentially paving the way for the complex and process-intense calculations forming the heart of any AI project.
In the year of its inception, McCarthy et al. defined AI as the problem of “making a machine behave in ways that would be called intelligent if a human were so behaving.”[1] However, in 1999, Gardner on the topic of AI put forward a more intricate definition highlighting “bio-psychological potential to process information” and its utility in the context of “culture.”[3] In a recent article, Kaplan and Haenlein. highlighted the practical considerations in AI as a system's ability to interpret external data correctly, to learn from such data, and to use those learnings to achieve specific goals and tasks through flexible adaptation.[4] Despite the varied representations, the thought surrounding AI consistently revolves around the concept of intelligent decision-making. Since any human decision-making process to be optimum, it requires inputs in terms of the situation or environment, the objectives, the resources, etc., and their processing in a very intricate internally interactive; it is safe to make a similar comparison to the thought of the machine. With experience, the human brain is trained to think more objectively in difficult situations. The concept of “experience” can be translated to a recursive process of feeding information, action, and error correction in terms of a machine's brain. It is this very context that makes a way for the popular subsets of AI research – machine learning and deep learning.
| From Machine Learning to Deep Learning: The Intelligence Perspective | | |
The advantage of a really intelligent machine is the availability of multiple probabilistic solutions instantaneously with an indication of the most likely solution. In public health, the systems and the interactions within and thereof are all highly nonlinear and nonnormative with multiplicity of factors contributing in and out of the dynamics under consideration, in the background of nonconvex optimization theory.[5] This dynamic absence of an equilibrium in a highly interactive environment sets up the very problem of time-efficient decision-making for the ultimate goal of the public good. The variations and interactions that constantly misbalance the systematic relations of health, health care, and the human are the externalities under consideration in this article in light of the role of machine learning and deep learning technology.
Machine learning is in fact an application of AI. It encompasses the use of algorithms in understanding the available information, i.e., data and analyzing it to arrive at an “intelligent” conclusion. In this context, it is important to understand that logic systems do not necessarily lead to intelligent conclusions. The difference can be understood in terms of tweaking in the logic programs allowing for weights or adaptations. Machine learning methods are designed to analyze and conclude from data with user-defined weights or conditions. The scope of machine created, i.e., “self-aware” weight or conditionality, however, is addressed in the domain of deep learning. Deep learning, thus, an extension of machine learning, learns from the data itself by repetitively correcting the errors and reaching to a more probabilistic conclusion.[5]
| Some Examples and Scopes | | |
The use of machine learning in public health has already begun and are scaling new heights. While the m-health initiatives handle a huge amount of patient/participant data and send simple messages or reminders in an automated manner, the database management systems in the different national programs are also improving from a simple data table to objectively programmed interactive data banks.[6] Crowdsourcing and media scanning is getting more and more used in health care.[7],[8],[9],[10] The validity and authenticity in part are also attributed to the machine learning programs built into the system. However, this requires constant access to the system and rigorous monitoring. A betterment can be achieved with the use of deep learning protocols where the program addressing the accuracy and authenticity of information from the public domain not only triangulates but also learns from the triangulation and rectifies the error instead of a programmer adjusting the sensitivity parameters. Another example in this context can be the automated vaccine stock maintenance procedures – the electronic Vaccine Information Network.[11] The development of the system lies dependent on the development in deep learning systems. The application of AI in radiology has evolved into a new horizon of radiomics, with the scopes of picture archive and computer systems still growing.[12] Naugler and Church have mentioned about the use of AI in laboratory diagnostics as well.[13] A more recent example should be considered regarding the uses of machine learning algorithms for predictions of the COVID19 pandemics, though accuracy issues still remain.[14]
The requirement of parameters in the program is further envisaged when we discuss the issue of analysis of big data. Unless the parameters are provided, the program is not able to perform the analysis as the algorithms are in most cases rooted in the parameters and variables despite them being nested in layers within the program. If we consider a system which collects data on patients suffering from different diseases and the medications provided, it can effectively provide suggestion to the physicians from its learning on which medication is likely to be more effective based on the patient profile.[15] Antimicrobial resistance in this context can be an issue for resolution. Rational computer-assisted antimicrobial prescribing can be further improved with the addition of data sources concerning the resistance patterns in areas. It can be considered under the generalized umbrella of an expert system for diagnosis and treatment.
| The Externalities: Threat or Caution? | | |
The factors that are not considered in the system usually cause the system to produce erroneous results. However, machine learning and deep learning algorithms thrive on minimizing this error and yielding the most probable results. Still, the question remains if the errors caused by the externalities in the system are too large the algorithms can fail.[16] To address this issue, more parameters and layers need to be added. While it appears a threat to the method, on the other hand, it brings in the much-needed caution statement that human verification is essential whenever felt required.[17] The whole process of precision public health is moving forward in the high-resource settings, riding on high-quality datasets, but in low-resource settings, the progress is still slow. While the high resource-settings can now think of progressing to deep learning surpassing machine learning algorithms, the resource-poor regions still need to properly set up the building blocks. The externalities would require to be handled by the human operators fixing the conditions and weights in the program sometimes empirically or intuitively.
The influences and interactions surrounding health and health care are innumerable. For instance, a hospital is set up in an area with an electronic medical record (EMR).[18],[19] After some days, it is observed that it is not being availed by the target population. While the EMR system will send an alert to the authorities regarding the underutilization of the hospital and may even provide a recommendation of shutdown based on cost–benefit, another algorithm can be produced that will scan the reasons of underutilization of the services in the area and identify the causes insisting intervention from the authority. The process of threat and caution can be complementary in this manner.
| The Way Forward | | |
Undoubtedly, applications of AI in public health have already brought about a paradigm shift in the provision of health care. With the global goal of universal health care, AI systems in public health can be considered even more important in the resource-poor underserved areas to make a systematic arrangement for health-care delivery. The primary health care is cardinal to achieve universal health coverage. The AI systems can help the resource-contained and the grass-root level settings with remote access, algorithm-driven diagnostic aids, notification regarding emerging threats, and automated analysis of the health data in defined regions. The scopes are even further with intensified AI research on diagnostic algorithms. These in particular will be of immense help where the physical presence of a physician can sometimes be challenging. AI and deep learning can also be integrated with the emerging trend of telemedicine, which can particularly be of immense help to provide the underserved with basic health care and allied decision-making.
The major hindrance, however, still lies in the fact that a machine can be taught the practice on the background of knowledge. The key components of attitude and behavior remain absent. The machine learning algorithms compensate for this with added layers of more complicated logic. When dealing with public health, the scanning of information, extraction, and analysis should always be done on the backbone of attitude, behavior, and beliefs. Although the algorithms can effectively address the externalities contributing to aberrations, they are unable to address the root. With growing endeavor and enthusiasm around AI and machine learning and its application in public health, addressing these issues for a stabilized system should be at the focus. Intuitively, the use of AI could decrease the cost of health care, though the evidence is still rudimentary.[20],[21] In conclusion, it is needless to say that the current affairs with machine learning and deep learning methodology must be maintained without any hiatus. The machines and programs should not incur debate over their usage in the health sector but over the method regarding the efficiency of the system.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
| References | | |
| 1. | |
| 2. | |
| 3. | Gardner HE. Intelligence Reframed: Multiple Intelligences for the 21 st Century. Hachette UK: Hachette UK; 2000.  |
| 4. | Kaplan A, Haenlein M. Siri, Siri, in my hand: Who's the fairest in the land? On the interpretations, illustrations, and implications of artificial intelligence. Bus Horiz 2019;62:15-25. |
| 5. | Sejnowski TJ. The unreasonable effectiveness of deep learning in artificial intelligence. Proc Natl Acad Sci U S A 2020;117:30033-8. |
| 6. | Istepanian RSH, Al-Anzi T. m-Health 2.0: New perspectives on mobile health, machine learning and big data analytics. Methods 2018;151:34-40. |
| 7. | Wazny K. Applications of crowdsourcing in health: an overview. J Glob Health 2018;8:010502. |
| 8. | Ye C, Coco J, Epishova A, Hajaj C, Bogardus H, Novak L, et al. A crowdsourcing framework for medical data sets. AMIA Summits Transl Sci Proc 2018;2018:273–80. |
| 9. | Groene EA, Mohammed I, Horvath K, Basta NE, Yared N, Kulasingam S. Online media scans: Applying systematic review techniques to assess statewide human papillomavirus vaccination activities. J Public Health Res 2019;8:1623. |
| 10. | Sharma R, Karad AB, Dash B, Dhariwal AC, Chauhan LS, Lal S. Media scanning and verification system as a supplemental tool to disease outbreak detection & reporting at National Centre for Disease Control, Delhi. J Commun Dis 2012;44:9-14. |
| 11. | |
| 12. | Deyer T, Doshi A. Application of artificial intelligence to radiology. Ann Transl Med 2019;7:230. |
| 13. | Naugler C, Church DL. Automation and artificial intelligence in the clinical laboratory. Crit Rev Clin Lab Sci 2019;56:98-110. |
| 14. | Lahiri A, Jha S, Bhattacharya S, Ray S, Chakraborty A. Effectiveness of preventive measures against COVID-19: A systematic review of in silico modeling studies in Indian context. Indian J Public Health 2020;64:156. |
| 15. | Ahuja AS. The impact of artificial intelligence in medicine on the future role of the physician. PeerJ 2019;7:e7702. |
| 16. | Osoba O, Welser W. An Intelligence in Our Image: The Risks of Bias and Errors in Artificial Intelligence. Santa Monica, Calif: RAND Corporation; 2017. |
| 17. | Wahl B, Cossy-Gantner A, Germann S, Schwalbe NR. Artificial intelligence (AI) and global health: how can AI contribute to health in resource-poor settings? BMJ Glob Health 2018;3:e000798. |
| 18. | |
| 19. | Lin WC, Chen JS, Chiang MF, Hribar MR. Applications of artificial intelligence to electronic health record data in ophthalmology. Transl Vis Sci Technol 2020;9:13. |
| 20. | Wolff J, Pauling J, Keck A, Baumbach J. Systematic Review of Economic Impact Studies of Artificial Intelligence in Health Care. J Med Internet Res 2020;22:e16866. |
| 21. | |
|
Search Pubmed for