Page 209 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

Chapter 7 Early detection and diagnosis using deep learning 199

DL in layman terms is the branch upheld by AI that has the
capability of initiating similar to the human brain; it provides
the machine with the ability of processing the data and helps
in creating the approximate patterns, which further helps in
decision-making. DL is also called deep neural network or deep
neural learning.
DL is a class that lies under ML, and it holds numeral layers
having the capability of extracting much higher level of features
in raw data. Information present in the environment is fetched
through different sensors; once fetched, it has to be converted
to defined data. Initially the raw data that we collect are a
completely unstructured data, and dealing with such kind of
data is really tough; hence, the model with higher capability is
required to deal with the situation. When it comes to application
in healthcare sector, the data can be in various forms ranging
from heartbeats to amount of viruses or infections that are found
during a urine test; with the vast variety of data, the requirement
of the complex system is necessary so that the prediction made or
the accuracy attained is fine enough and it does not result in an
underfitting model.
In cases of disease prediction such as diabetes or cancer, the
model obtained should be an overfitting one because a tightly
fitted model helps us in better prediction and simultaneously
results in a highly accurate model. Reason of obtaining such
models is that we cannot take any chances when it comes to
the lives of people. Hence, prediction made should be highly
accurate so that doctors can then suggest better medication.
Now while incorporating DL for diagnostics, we use different
kinds of neural networks such as DNN (deep neural network),
RNN (recurrent neural network), CNN (convolutional neural
network), and so on for dealing with varied real-life problems.
A number of researches were held and are still going on, which
include DL in dealing with healthcare technologies of biometrics,
medical image processing, and much more. DNN is a type of
neural network defining inputs and outputs with the complex
layers computing and building blocks that include processes
such as transformation and nonlinear functions [3]
DL has the capability of solving problems, which are almost
impossible to solve otherwise with traditional AI [4]. The sub-
missions of DL can lead to malevolent activities, but the
constructive usage of the knowledge is considerably wider.
Back in the year 2015, it was renowned that DL has a strong
pathway on the way to functioning with huge data sets, and
therefore, applications of DL areprobabletobemuchmore
wider in the coming future [4].