IEEE Systems, Man and Cybernetics Magazine - July 2022 - 30

development or the predictions obtained by handcrafted
conventional ML and DL methods [19]. AI has no voluntary
authority, and for this reason, medical experts
believe that it cannot be said to have a character or
motive. Moreover, so-called black boxes of different AI
technologies, such as handcrafted
conventional ML and DL methods,
are another
signi ficant
reason why medical experts do
not trust AI-based medical systems
[19]. Hence, medical experts
are interested in understanding
the performance of handcrafted
conventional ML and DL methods
under uncertainty.
One vital strategy to increase
Another important research domain is the application
The remarkable
efficiency of AI
has paved the way
for other branches
of science and
trust in these models is applying
uncertainty measurement and
quantification methods. The growing
importance of the application
of AI-based techniques is visible.
But more important, uncertainty
measurement and quantification of these AI methods in
various applications have become a key issue in recent
years. However, this perspective highlights only the main
important reasons for uncertainty measurement and
quantification of ML and DL methods used in medical
data analysis.
Applications of ML/DL
Nonmedical Applications
In this section, we discuss nonmedical and medical applications
and briefly present important AI methods in
real applications. As mentioned earlier, AI methods are
used in various applications. For example, researchers
and countries have highly regarded climate change as a
significant change in regional or global climate patterns.
To predict and track noticeable climate changes,
AI methods have been used significantly [21]. The
results reveal the critical effects of these methods to
prevent cl imate change or reduce its destructive
effects. But AI applications are not limited to controlling
climate change. Self-driving car (also called AV or
driver-less car) technology is another critical and sensitive
application of AI methods. To secure the route of a
self-driving car from the first point A to target point B,
divers use AI methods [22]. On the other hand, driver
behavior analysis can be listed as one of the essential
research domains to improve the quality of transportation
and reduce the number of deaths caused by accidents.
Hence, monitoring drivers' behavior seems to be
of great importance. Determining risky driving behaviors
helps to improve road safety. Thus, AI methods
have indicated promising achievements in detecting
risky driving behaviors [23].
30 IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE July 2022
engineering to apply
it in their fields
extensively.
of different natural language processing (NLP), SA, text
mining, emotion analysis, and so on [24], [25]. The
increasing advancement in computer technology has led
to many texts with various subjects. NLP should ultimately
exploit meaning (called
semantics) from the text. In addition,
AI methods have demonstrated
promising results in text
analysis [26]. Along with all these
applications, there are several
other important applications of AI
methods; for example, handcrafted
conventional ML and DL demonstrated
stunning performance
in the area of weather forecasting
[27], online marketing [28], energy-related
studies [29], and natural
disaster modeling [30]. Finally,
business analytics has significantly
exploited the use of AI
methods [31]. Such intelligent
techniques have inspired business people and developers
to create AI models to yield accurate decisions.
The Era of ML/DL for Medical Data Analysis
Ever since its inception, AI has been widely used in many
applications. Today, many automated systems are proposed
to detect and predict various diseases accurately
in the health-care domain (e.g., breast cancer [32], prostate
cancer [33], liver cancer [34], cancer cells [35], cardiovascular
disease [36], COVID-19 [37], and so on).
Recent statistics reveal that in the United States, approximately
only 5% of outpatients receive the wrong diagnosis
yearly [38]-[40]. However, these diagnostic errors are
not limited to outpatients but are common, especially
when detecting/diagnosing patients with severe medical
conditions. Accordingly, estimates show that approximately
20% of patients with serious medical conditions
are misdiagnosed at primary care levels [41]. As a result
of these errors, one out of three misdiagnoses can cause
serious harm to patients, leading to harmful clinical consequences
[42].
These statistics indicate a large number of errors in
medical diagnoses. This reveals that errors may even
occur in physicians' diagnoses. Therefore, alternative
solutions are strongly recommended to reduce these
errors. Evidence shows that medical experts foresee AI
IEEE Systems, Man and Cybernetics Magazine - July 2022

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - July 2022
