Page 289 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

280 Chapter 10 Deep neural network in medical image processing

identifying whether there is a person or a car in the picture or
there is an animal or flower in the picture. Various supervised
and semisupervised algorithms are deployed for this purpose;
algorithms are run against partially or fully tagged image data
sets to achieve this step.
Extracting knowledge: This is the final frontier in image pro-
cessing where systems can generate quantifiable intelligence
like in a driverless car where various sensors and cameras are
used to get actionable intelligence based on which the machine
decides the next move.

2.6 Machine learning and its types
Machine learning is the study of mathematical and statistical
models and algorithms associated with them to perform a prede-
fined task with explicit instructions for the same. In Layman terms,
machine learning is a field of study which empowers the machines
by providing them the ability to learn from data and improve their
functioning from experience without being explicitly programmed
for the same, much like humans train themselves. Machine learning
primarily focuses on development of computer systems that can ac-
cess data and use those data to further improve themselves. The pri-
mary aim of this type of program is to look for common patterns in
the data and classify them according to various parameters and
make better decisions based on those classifications.
There are primarily three types of machine learning algorithms:
(Fig. 10.4)
• Supervised learning
• Unsupervised learning
• Reinforcement learning
But before diving into types of machine learning algorithms,
we have to understand two very important terms.

Figure 10.4 Machine learning types.