Basic

Medical
Image
Processing
and Analysis
 Medical Image Processing

•Deals with the development of problem

specific approaches to enhance the raw
medical data for the purpose of selective
visualization as well as further analysis

Processing
vs Analysis Medical Image Analysis

•Concentrates on the development of

techniques to supplement the usually
qualitative and subjective assessment of
medical images by human experts.
 • Medical Images are
generated by using image
reconstructions or
Medical tomography, techniques with
data from projection that are
Image collected from detector
coupled with various energy
Processin •
sources.
These energies can be from x
g -ray, single photon emission,
positron emission, ultrasound,
nuclear magnetic resonance
or light photon.
 Processing parameters and image
manipulation is the same for both system.

Medical
Image Preprocessing takes place in the computer
where the image histogram is determined

Processing
Postprocessing is done by the technologist
though various user function.
 Images are classified
based on their form and
method used to produce
them.
Image
Formation
and It can either be
classified into visible,
Representatio non visible, optical and
mathematical images.
n An
alo
gi
ma
Mathematical ges
images is an
important in dD
digital imaging. igit
al
im
ag
 Spatial Domain

•This represents the anatomical or

physiological functions in space.

Image
Domains
Frequency Domain

•This represents the mathematical

transformation of the anatomy.
•Fourier transform pair
•Discrete fourier transform pair
 Quality Evaluation

• Plays an important role in assessing any

new hardware, software, image acquisition
techniques, image reconstruction or post-
processing algorithms.
Statistical Analysis and Entropy

• Statistics of medical imaging fills the gap

and provides a theoretical framework for
statistical investigation into medical
imaging technologies.
 Coding and Decoding

• Color information in digital images is used

for many practical purpose, mainly in
security fileds and also with gray scale
imaging applications.
 Stable and Unstable Signal
Processing
• The system is said to be stable only when
the output is bounded for bounded input.
For a bounded input, i the output is
unbounced in the system then it is said to
be Unstable.
• A stable system satisfies the BIBO.
• (Bounded input for Bounded outpu-t)
means finite in amplitude.
 Transformation

• An image transformation can be applied to

an image to convert it from one domain to
another. Viewing an image in domains
such as frequency or hough space
enables the identification of features that
may not be easily detected in the spatial
domain.
Common Image Transforamations
• Hough Transform
• Radon Transform
• Discrete Cosine Transform
• Disctrete Fourier Transform
• Wavelet Transform
 Hough Transform

• The Hough transform is a feature

extraction technique used in image
anaysis, computer vision, and digital
image processing.
• The purpose of the technique is to find
imperfect instances of objects within a
certain class of shapes by a voting
procedure.
 Radon Transform

• Radon transform is an integral transform

whose inverse is used to reconstruct
images from medical CT scans.
 Discrete Cosine Transform

• Discrete cosine transform expresses a

finite sequence of data points in terms of
sum of cosine functions oscillating at
different frequencies.
 Disctrete Fourier Transform

• Discrete fourier transform is of paramount

importance in all areas of digital signal
processing. It used to derived a frequency-
domain (spectral) representation of the
signal.
 Wavelet Transform

• A mathematical technique which can

decomposes a signal into multiple lower
resolution levels by controlling the scaling
and shifting factors of a single wavelet
function.
 Image Background
Reconstruction Removal
Pre-
processin
g Noise Removal
Image
Compression
 • It is the process based
on the use of an
algorithm, referred to as
a reconstruction
algorithm.
Image – 2D Fourier Projection
Reconstruction Theorem
– Algebraic
Reconstruction
Method
 2D Fourier Reconstruction
Theorem
• This is the reconstruction of 2D images
coming from 1D projections.
• 180 degree projection with 1 degree
increment is necessary to produce an
image with satisfactory quality.
Algebraic Reconstruction Method
• This is used when there are incomplete
number of projections for the
reconstruction of the images.
 Background
Removal
• Images obtained by projection
radiography contain unexposed areas
due to X-ray collimation
• In digital images, unexposed areas
appearing white on a display monitor is
defined as background in this context
 Noise suppression or
denoising

Noise is always present in all

Noise medical images.

Removal There are several ways to

reduce the amount of noise
present in the images.

Using filters would also

contribute in reducing the
amount of noise present in
the image.
 Averaging or Block Filtering

•Simplest filter
•Each pixel is replaced by the average of its
neighbor

Noise Gaussian Filter

removal
•It blurs the image

Median Filtering

•Simple but efficient in noise reduction with

good performance
 Anisotropic Diffusion Filtering

•smooths regions without strong edges

while preserving edges.

Noise Bilateral Filtering

Removal •It was developed similar to gaussian

filter.

Nonlocal filter

•newer technique that examines a

‘‘patch’’ centered about each pixel
Image Compression

It serves to reduce the size of the image in order to decrease

transmission time and reduce storage space.

Lossy compression
Two forms of Image compression:
Lossless compression

Wavelet (special wavefor

New form of compression
m ) compression
 Filtering

Post- Enhancement

Medical
Image
Processing
Registration

Classification, Texturing and

Segmentation
Filtering
This is a processing method that enhances or removes a specific
component in a signal or image.
 It is the process of aligning images with one
another

Registratio It may be intra-subject, intersubject,

n intramodality and inter modality.

It uses algorithms that are classified by

transforms allowed, similarity metric and
minimization.
 Segmentation

It is the separation of an image into meaningful components.

This is an essential step if the Thresholding

user wants to perform automated
Region based methods
measurements on organs or
diseases. Edge Based Methods
 es
da
ta
fro
m
tw
o
It is the assignment ofor
a
meaningful name to a m
or
group of pixel or voxels.
ei
Classificati m
ag

on es
of
th
Multispectral es
Classification a
m
e
an
at
o
m
yt
o
 • It serve to generate an
image that is more
pleasing to the observer.
Image • This include contrast
Enhancemen enhancement, edge
t enhancement, spatial
and frequency filtering,
image combing and
noise reduction.
 Spatial frequency filtering

Edge Enhancement

•Occurs when fewer pixels in the

Image neighborhood are included in signal
average.
Enhancemen •The smaller the neighborhood, the
greater the enhancement
t •High pass filter is used when known
frequencies can be amplified and others
can be supressed.
•Suppresing frequencies is aka as
masking.
Rx: Loss of small details
 Image enhancement

• Smoothing
– Aka as low-pass filter
– Averaging each pixels frequency with surrounding
pixel value to remove high frequency noise.
– Reduction of noise and contrast.
– Useful for viewing small structures such as fine
bone tissue.
 Also referred to as contrast scaling

Contrast
Enhancemen
t
It is used to optimize the image contrast and
density to enhance diagnostic interpretation of
the image.
 • Image Display
– Soft copy viewing
refers to the display
of the digital images
Medical Image at a computer
Displaying, workstation/ display
Visualization monitors
and
– Hard copy refers to
Representation
viewing images on a
film or another
physical medium.
 • Display Monitors/Workstations
– It consist of hardware and
software to facilitate the
display of digital images for
Medical diagnostic inter pretation
Image and for review purposes.
Displaying, – Cathode Ray Tube
Visualization • It displays an image by
and accelerating and
focusing electrons to
Representatio strike the faceplate
n – Liquid Crystal Display
• It passes light through
liquid crystals to display
the image on the glass
plate.
 Medical Image Displaying,
Visualization and Representation
 • There are 4 types of
workstation that are
Medical available in a PACS
Image environment
Displaying, – High-resolution
Visualization display workstation
and – Medium-resolution
display workstation
Representatio
– Desktop workstation
n
– Hard-copy
workstation
 Medical
Image
• Display workstation also have
image processing functions.
– It allows the user to

Displaying, manipulate or adjust the

image to suit their viewing
needs.

Visualization • Subtraction
• Contrast

and
Enhancement
• Edge enhancement
• Black/White

Representatio Reversal
• Smoothing

n
3D Volume Rendering
and Ray – Tracing
• 3D visualization
– Medical image modalities such as CT and MRI
generate 3d images.
– Postprocessing techniques including volume
rendering allow these to be displayed.
• Volume Rendering this is a visualization
technique to display a 2D image of a 3D data
set that retains access to the original voxel
data.
 Volume • Projection Rendering
renderin • Surface Rendering
• Direct Volume Rendering
g
• Multi-planar Reconstruction (MPR)
– This reconstruction method allows
real time viewers to slide through a
given volume in any direction.
– It creates other sectional images form
axial data sets. Projectio
n Renderi
– It provides better diagnostic
information
• Maximum Intensity Projection
– Projection of voxels with high
intensity.
ng
• Minimum Intensity Projection
– Projection of voxels with lowest
intensity.
Surface Rendering
• Also known as shaded surface display
• Surface-rendered image that provides a
realistically looking three-dimensional view of
the surface of a structure of interest within the
acquired volume set.
 • It renders images of 3D
volume data sets directly
Direct from the 3D volume data
without extracting any
Volume geometrical information
about the objects
Rendering captured.