Page 61 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

Chapter 2 Deep convolutional neural network in medical image processing 49

4.4 Chest
The most commonly addressed issues in thoracic image anal-
ysis are characterization, detection, and classification of nodules.
These applications are for computed tomography as well as radi-
ography. Several works related to this have been summarized in
Table 2.2. Many studies give the addition of features derived
from DL networks to the already existing features or even
compare CNN with the traditional ML methods taking the manu-
ally constructed features. In many works, multiple diseases are
detected with a single system using chest X-rays. In CT, a very
popular research area is the detection of textual patterns that indi-
cate interstitial lung diseases. One of the most common radiology
examinations is chest radiography with the help of which many
studies are carried out. In many of the work, a big set of images
having text information are used for the purpose of system
training that combines RNN for text processing and CNN for im-
age processing. This is an area of research that is expected to be
flourishing soon.

4.5 Cardiac
Many aspects of cardiac image analysis have applications of
DL. The summarization of different works related to cardiac im-
age analysis is done in Table 2.2. Various applications include
tracking, segmentation, slice classification, automated calcium
scoring, image quality assessment, coronary centerline tracking,
and superresolution. Among these applications, left ventricle seg-
mentation is one of the most common tasks. Most of the studies
are done on MRI image modality. It is expected that more
image-based computer-aided detection and diagnostic systems
with further validation studies and technology transfer efforts
can be put into clinical use in the near term.

4.6 Abdomen
Several papers based on abdominal image analysis addressed
the problem of localization and segmentation of organs. Most of
the works done are based on MRI, CT, or endoscopic images. MRI
modality is generally used for the analysis of the prostate,
whereas CT is used for examining the liver, kidney, pancreas,
and bladder. Endoscopy is one of the modalities that are very
recently being used for the internal examination of the stomach
and intestine. Summarization of various studies done related to
the abdominal organs is done in Table 2.2. With the increase of