|Year : 2019 | Volume
| Issue : 2 | Page : 95-97
Machine learning in medicine: Will doctors meet their waterloo?
Department of Community Medicine, Dr. DY Patil Medical College, Hospital and Research Centre, Dr. DY Patil Vidyapeeth, Pune, Maharashtra, India
|Date of Web Publication||25-Mar-2019|
Department of Community Medicine, Dr. DY Patil Medical College, Hospital and Research Centre, Dr. DY Patil Vidyapeeth, Pune, Maharashtra
Source of Support: None, Conflict of Interest: None
|How to cite this article:|
Banerjee A. Machine learning in medicine: Will doctors meet their waterloo?. Med J DY Patil Vidyapeeth 2019;12:95-7
The concept of artificial intelligence (AI) was first proposed by a group of scholars cutting across disciplines in 1956. John McCarthy, a mathematician, who headed the group, conceptualized that “…every aspect of learning or any other feature of intelligence can in principle be so precisely defined that a machine can be made to simulate it.” This seminal concept spawned AI and subsequently its offspring, machine learning. Over the decades, faster computers with high capacity and interconnectivity, combined with quantification of predictors and outcomes, and pattern recognition have enabled application of AI and machine learning in medicine exponentially. So much so that some believe that AI will replace 80% of physicians in coming decades.
Offshoots of AI and machine learning in medicine are P4 (predictive, preventive, personalized, and participatory), and stratified medicine whereby patients are classified on their disease subtypes, risk, prognosis, or response to drugs using “in silico” methods., These emerging fields are deeply related and dependent on data science specifically AI or machine learning hence the term “in silico” (alluding to the silicon in computer chips).
Large data sets, data mining, and “deep learning” have generated interest into machine learning applications in medicine particularly in disciplines such as radiology,, histopathology, and personalized medicine.,,, Diagnostic accuracy of machines has at times been better than specialist doctors in some instances.
The good old family doctor in general practice got marginalized with specialization and subspecialization. Is it the turn of specialists to meet their Waterloo? The most vulnerable appear to be the radiologists. Will machine learning knock radiology off its perch as medicine's most coveted discipline? This apprehension is illustrated by an e-mail received by a professor of radiology at Stanford University from one of his students expressing concern that radiology is not a viable profession anymore.
The debate promises to be protracted. AI can simulate the attributes of human intelligence such as learning, logical thinking, autocorrection, and pattern recognition. Clinicians of yesteryears prided themselves to be good observers. The patient was under scrutiny from the moment of crossing the threshold to the doctor's clinic. The gait, the facial expression, the speech, the grip, and other subtle cues provided vital leads to the astute clinician. This power of observation often enabled the doctor with years of experience in pattern recognition to clinch a diagnosis even before the patient narrated his medical history.
The classic example is Dr. Joseph Bell, a clinician with amazing powers of observation, who inspired Sir Arthur Conan Doyle to create the fictional detective character Sherlock Holmes. Can machines replicate the subtle observation skills of such clinicians?
Sebastian Thrun, a former Professor in Computer Science at Stanford University, believes so. He predicts that constant observation with the help of machine learning to detect signs of early disease is possible. Our phones can be made to detect subtle changes in our speech to identify early Alzheimer's disease. The steering wheels in our cars can be trained to detect early Parkinson's disease by the way we handle them. Our bathrooms can be scanning zones with the aid of benign ultrasound or magnetic resonance to detect early tumor in our internal organs.
Thrun believes that machine learning will be an aid to doctors and will not replace them; these tools will augment the professional skills of doctors. The phone did not replace the human voice. It only served as an augmentation device. Similarly, he believes that machine learning will serve to augment the capacity of the human mind, offering expertise and assistance to doctors rather than replacing them.
Geoffrey Hinton, a computer scientist at the University of Toronto, and a descendant of George Boole who invented Boolean algebra, the foundation of computing, is less kind to doctors, particularly radiologists. He predicts that in the next 5 years, the machine will outperform radiologists and goes to the extent of stating that they should stop training radiologists. Hinton does not spare other medical disciplines either. He prophesizes that deep learning machine algorithms would make pathological diagnoses, read Pap smear More Detailss, interpret heart sounds, diagnose skin lesions, and also give prognosis in psychiatric patients.
Jorg Goldhahn, from the Institute of Translational Medicine, Zurich, Switzerland, is equally harsh on doctors. He is of the opinion that while machines may not always be superior to doctors presently, the issue is technical rather than basic. With deep learning which is part of machine learning which continuously corrects and upgrades itself faster than humans, AI will surpass doctors in diagnosis and treatment in times to come. In addition to diagnosis and therapeutics, machines will be able to perform complex clinical tasks including ethical and cost–benefit analysis. He argues that it is beyond the capacity of physicians to master the exponentially increasing medical knowledge, and at the same time, care for their patients because of the sheer volume of data. Machine learning will have an edge; natural language processing will help them incorporate the rapidly increasing medical literature and also predict complex pathways for instance about adverse drug reactions. Goldhahn concludes that while machine learning may not be perfect today, in due course of time, they will outperform doctors.
Others differ. The contrary position is that doctors will always be better in caring for the patient as a person for which soft skills and insight on social and cultural background is essential. Humans have the empathy which involves trust, respect, courage, and accountability which cannot be replicated by machines. Machines will also face a dead end when confronted with emotional, social, and nonmeasurable issues that lead to sickness. Patients may still want a doctor to listen to them and whom they can trust. Moreover, the healing effect of a good doctor–patient relationship, tailored to the cultural and social background of the patients, is well established irrespective of the efficacy of the prescribed drugs. Does the writing on the wall indicate rise of Social Medicine as a discipline? Do machines find the concepts of Social Medicine too fuzzy too grasp? So do many in the profession, including Social Medicine specialists!
Besides, posing a challenge to doctors, machine learning is filling up vacant posts of absent doctors in remote and resource-poor settings where it holds great potential for transforming the health-care services. Nearer home, the Health Minister of Uttar Pradesh, a state in India known for its poor health indicators, announced that 500 Primary Health Centers in the state will be manned by robotic machines in a phased manner to meet the health needs of the population.
So in future will we have driverless cars and doctorless hospitals and health centers? The debate goes on…why don't we allow AI and machine learning to answer this question?
| References|| |
Fröhlich H, Balling R, Beerenwinkel N, Kohlbacher O, Kumar S, Lengauer T, et al.
From hype to reality: Data science enabling personalized medicine. BMC Med 2018;16:150.
Flores M, Glusman G, Brogaard K, Price ND, Hood L. P4 medicine: How systems medicine will transform the healthcare sector and society. Per Med 2013;10:565-76.
Lee JG, Jun S, Cho YW, Lee H, Kim GB, Seo JB, et al.
Deep learning in medical imaging: General overview. Korean J Radiol 2017;18:570-84.
Shen D, Wu G, Suk HI. Deep learning in medical image analysis. Annu Rev Biomed Eng 2017;19:221-48.
Djuric U, Zadeh G, Aldape K, Diamandis P. Precision histology: How deep learning is poised to revitalize histomorphology for personalized cancer care. NPJ Precis Oncol 2017;1:22.
Miotto R, Li L, Kidd BA, Dudley JT. Deep patient: An unsupervised representation to predict the future of patients from the electronic health records. Sci Rep 2016;6:26094.
Beaulieu-Jones BK, Orzechowski P, Moore JH. Mapping patient trajectories using longitudinal extraction and deep learning in the MIMIC-III critical care database. Pac Symp Biocomput 2018;23:123-32.
Choi E, Schuetz A, Stewart WF, Sun J. Using recurrent neural network models for early detection of heart failure onset. J Am Med Inform Assoc 2017;24:361-70.
Choi E, Bahadori MT, Schuetz A, Stewart WF, Sun J. Doctor AI: Predicting clinical events via recurrent neural networks. JMLR Workshop Conf Proc 2016;56:301-18.
Yu KH, Zhang C, Berry GJ, Altman RB, Ré C, Rubin DL, et al.
Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun 2016;7:12474.
Goldhahn J, Rampton V, Spinas GA. Could artificial intelligence make doctors obsolete? BMJ 2018;363:k4563.
Osborn J. Observation, Sherlock Holmes, and evidence based medicine. Med Secoli 2002;14:515-27.
Lim S, Lee K, Kang J. Drug drug interaction extraction from the literature using a recursive neural network. PLoS One 2018;13:e0190926.
Gawande A. “Tell me where it hurts.” New York. The New Yorker; 2017. p. 36-45.
Di Blasi Z, Harkness E, Ernst E, Georgiou A, Kleijnen J. Influence of context effects on health outcomes: A systematic review. Lancet 2001;357:757-62.
Stonington SD, Holmes SM, Hansen H, Greene JA, Wailoo KA, Malina D, et al.
Case studies in social medicine – Attending to structural forces in clinical practice. N Engl J Med 2018;379:1958-61.
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.