Artificial Intelligence in healthcare

The complexity and growth of data in the healthcare sector means that Artificial Intelligence
(AI) is being used more and more in this area. Various types of artificial intelligence are
already used by customers and service providers, as are life sciences companies. The most
important application categories include diagnostic and treatment recommendations, patient
participation and compliance, and administrative activities. Although there are many cases
where AI can perform healthcare tasks as well or better than humans, implementation factors
will prevent extensive automation of healthcare professions over a considerable period of
time.

In recent years, machines have surpassed human performance in many cognitive tasks. The
transformative power of artificial intelligence (AI) extends too many industries. The effects
of AI in healthcare were very promising and could completely transform healthcare in the
near future. AI can be used in many health related areas, from hospital care and clinical
research to drug discovery and diagnosis prediction. The rapidly increasing availability and
low costs of high-performance computing resources are leading to the digital transformation
of the healthcare system. The use of innovative technologies in daily medical practice enables
secure, real-time access to data and big data analytics. This increases collaboration between
specialists and improves the overall quality of treatment. Large organizations use big data
analytics to diagnose disease. For example, IBM's Watson for Health helps healthcare
organizations analyze large amounts of health-related data to improve diagnosis [51]. An
obstacle to data analysis is the heterogeneity of medical information, e.g. medical journals,
symptoms, test results, treatment cases. Therefore, big data technology used with novel
artificial intelligence methods should provide doctors with diagnostic tools. Watson may
review, store and process the medical data mentioned above. Another example of successful
collaboration between technological innovators and medical institutions is Google's Deep
Mind Health [52]. Researchers and clinicians work with patients to solve real-world health
problems by using machine learning (ML) algorithms, such as neural network models, that
mimic the human brain. ML looks for hidden patterns in the data to identify patients at risk,
segment regions of interest, evaluate data, diagnose and make a decision [53,54].

[51] Ibm is counting on its bet on watson, and paying big money for it, (2016)
https://www.nytimes.com/2016/10/17/technology/ibm-is-counting-on-its-bet-on-watson-and-
paying-big-money-for-it.html, Accessed 7th Sep 2020
[52]Deep mind health, (2019), https://deepmind.com/applied/deepmind-health/, Accessed 7th
Sep 2020
[53] W.-H. Hu, D.-H. Tang, J. Teng, S. Said, R. Rohrmann, et al. Structural health
monitoring of a prestressed concrete bridge based on statistical pattern recognition of
continuous dynamic measurements over 14 years, Sensors, 18 (12) (2018), p. 4117
[54] J. Yang, G. Sha, Y. Zhou, G. Wang, B. Zheng, Statistical pattern recognition for
structural health monitoring using esn feature extraction method, Int J Robot Autom 33 (6).

AI in medical image analysis

In recent decades, medical imaging has become an integral part of medical care. Images were
widely used for the detection, verification, differential diagnosis and treatment of diseases
and in rehabilitation. The AI algorithms achieved significant results when processed. Doctors
analyze various digital medical imaging modalities, including X-ray, ultrasound (US),
computed tomography (CT), magnetic resonance imaging (MRI), and positron emission
tomography (PET), mammography, retinal photography, histology, morphology and
dermoscopy slides. Table 2 summarizes the data on the imaging modalities and their most
common uses. Reporting images is a time-consuming task and is performed primarily by
experienced radiologists and physicians. Image reading is subject to error due to variation in
visual appearance of pathology and approaches to interpreting images. The potential fatigue
of human experts can also be responsible for an incorrect diagnostic decision. For example,
the sensitivity and specificity of mammography examinations were reported to be between
77-87% and 89–97% respectively [55]. 
[55] M.S. Bae, W.K. Moon, J.M. Chang, H.R. Koo, W.H. Kim, N. Cho, A. Yi, B. La
Yun, S.H. Lee, M.Y. Kim, et al., Breast cancer detected with screening us: reasons for
nondetection at mammography, Radiology, 270 (2) (2014), pp. 369-377

Patient engagement and adherence application

Patient commitment and adherence have long been considered the “last mile” problem of
healthcare, the final barrier between poor and good health outcomes. The more patients are
proactively involved in their own well-being and care, the better the results - utilization,
financial results, and member experience. Big data and artificial intelligence are increasingly
addressing these factors. Healthcare providers and hospitals often use their clinical
experience to develop a plan of care that they know will improve the health of an acute or
chronic patient. However, this often does not matter if the patient does not make the
necessary behavioural adjustments, e.g. B. Lose weight, make a follow-up appointment, fill
prescriptions, or follow a treatment plan. Non-compliance, when a patient does not follow
treatment or does not take prescribed medications as recommended, is a major problem. In a
survey of more than 300 clinical and healthcare leaders, more than 70% of respondents said
that less than 50% of their patients were highly engaged, and 42% of respondents said that
less than 25% of their patients were highly engaged [56]. If Greater patient involvement leads
to better health outcomes, can AI-based skills be effective in personalizing and
contextualizing care? There is a growing emphasis on using business rules engines and
machine learning to drive nuanced interventions throughout the continuum of care [57].
News alerts and specific, relevant content that triggers action at critical times is a promising
research area. Another growing focus in health care is the effective design of “electoral
architecture” to shape patient behaviour in a more prospective way based on the findings of
practice. Using information provided by EHR systems, biosensors, watches, smartphones,
chat interfaces, and other tools, the software can tailor recommendations by comparing
patient data with other effective treatment routes for similar cohorts. Recommendations can
be shared with care providers, patients, nurses, call center agents, or care delivery
coordinators.

[56]. Davenport TH, Hongsermeier T, Mc Cord KA. Using AI to improve electronic health

records. Harvard Business Review 2018. https://hbr.org/2018/12/using-ai-to-improve--
electronic-health-records. 
[57] . Volpp K, Mohta S. Improved engagement leads to better outcomes, but better tools
are needed. Insights Report. NEJM Catalyst, 2016, https://catalyst.nejm.org/patient-
engagement-report-improved-engagement-leads-better-outcomes-better-tools-needed. 
5. AI in precision medicine

Precision medicine is an emerging field for the prevention and treatment of disease. It takes
into account individual variability in genes, environment and lifestyle. In recent years, the
healthcare paradigm has changed [58,59]. The field of precision medicine has advanced
rapidly due to the development of AI algorithms that could analyze large amounts of genomic
data to predict and prevent disease. Traditional medicine applies uniform treatment to the
entire population, while precision medicine develops personalized treatment regimens for
subgroups of patients. Some factors may be more important to a particular subgroup. This
motivates clinicians and medical researchers to develop new approaches to subgroup
identification and analysis. This is an effective strategy for personalized treatment [60]. The
original concept of precision medicine included prevention and treatment strategies. These
strategies take individual variability into account by evaluating large data sets that include
patient information, medical images, and genomic sequences [61]. This approach allows
clinicians and researchers to predict which treatment and prevention strategy will work.

[58] F.S. Collins, H. Varmus, A new initiative on precision medicine, N Engl J

Med, 372 (9) (2015), pp. 793-795

[59] Z.-G. Wang, L. Zhang, W.-J. Zhao Definition and application of precision medicine,

Chin J Traumatol, 19 (5) (2016), p. 249
[60] M.Z. Nezhad, D. Zhu, N. Sadati, K. Yang, P. Levi, Subic: a supervised bi-clustering
approach for precision medicine, 2017 16th IEEE International conference on machine
learning and applications ICMLA, IEEE (2017), pp. 755-760
[61] J.-G. Lee, S. Jun, Y.-W. Cho, H. Lee, G.B. Kim, J.B. Seo, N. Kim, Deep learning in
medical imaging: general overview, Korean J Radiol, 18 (4) (2017), pp. 570-584
 [DO PROPER FORMATING OF FOLLOWING TABLE]
Table 2. Review of image modalities and their common application.

Modality Data Method Targeted Acquired information Spheres of medical

dimentionality structure diagnostics
Radiography 2D Using X-rays to Radiodense X-ray density Muscular-skeletal
syn.: X-ray view an internal (e.g. bone) and radiology (traumas,
structure of a object radiolucent bone pathology,
(e.g., air filled) orthopaedics), dentistry,
structures pulmonology, express
abdominal
Mammography A special type of X- Breast masses, Breast cancer detection
ray radiography calcifications
Ultrasound 2D-4D Examining the sound Soft tissues Reflection, refraction 1.
waves propagation and attenuation of a Diseases of
through tissues sound wave inner organs

2.
Functional
diagnostics in
cardiology,
gastroenterology

3.
Perinatology

CT 3D Reconstructing Radiodense X-ray density of Neurology (e.g.,

images from the and tissues intracranial
signals acquired radiolucent hemorrhages),
after passing an X- internal pulmonology, oncology
ray beam through the structures traumatology and
body which look orthopaedics
contrast
compared to
other objects
MRI 3D Studying magnetic Soft tissues Magnetic features of Neurology (e.g.,
features of tissues by masses, fluids, cells ischemic stroke),
employing the effect and tissues vertebrology, oncology,
of nuclear magnetic muscular-skeletal
resonance radiology (joints,
tendons, bone marrow),
cardiology (heart MRI),
internal medicine,
urology, gynaecology,
perinatology (fetal
MRI),
Contrast
Modality Data Method Targeted Acquired information Spheres of medical
dimentionality structure diagnostics
 - X-ray 2D Measuring an The integrity 1. 1.
accumulation of a of walls, the Contrast Anatomic
 - US 2D-4D
contrast media inside shape of passage (eg, abnormalities
 - CT 2D-4D of the lumen an hollow organs, irrigoscopy, (eg, an
organ or studying its the uteroscophy) obstruction,
passage through a vascularization mass lesion,
tract of parenchyma 2. decay)
Contrast
accumulation 2.
(eg, Pathologic
mammography) vascularization
because of a
3. tumor,
Vascularization inflammation
(eg,
angiography) 3.
Vessel
pathology (an
obstruction,
bleeding),
angiogenesis

 - MRI 2D-4D

PET 3D Changes at the Structures An accumulation of Oncology, neurology

cellular level with metabolic specific tracers may (neurodegeneration),
measured with abnormalities reflect etc.
SPECT 3D radiotracers or abnormal metabolic Oncology, cardiology,
radiopharmaceutical activity) functional studies
s

Microscopy 2D An optical A Visual appearance of a Confocal microscopy in

magnification for the microanatomy surface dermatology
visualization of of cells,
structures tissues and
organs
Cytology, 2D An accumulation of Pathologies may Laboratory diagnostics,
histology specific staines by account for chemical pathology
tissues shift and other
microstructural
changes justified by
staining
Endoscopy 2D An examination of Internal An observation of Gastroenterology,
the interior of a cavities and internal surfaces and endobronchial studies,
Modality Data Method Targeted Acquired information Spheres of medical
dimentionality structure diagnostics
hollow organ or hollow organs lumens thoracic and abdominal
cavity of the body surgery, urology
Funduscopy 2D An eye examination Central and Visual appearance of
with a special lens peripheral retina, optic disc,
retina, optic vitreous humor
disc and
macula