Mohammad Dawood - Computer Vision and Pattern Recognition

Report
Medical Imaging
Mohammad Dawood
Department of Computer Science
University of Münster
Germany
Medical Imaging, SS-2011
What is medical imaging?
Medical imaging is the process of acquiring
images without or with minimal invasion for the purpose of
detecting, diagnosing, quantifying or treating a disease.
Techniques and methods from image processing are used
to assist the clinicians.
Mohammad Dawood
2
Medical Imaging, SS-2011
Structure of the Course
1.
2.
3.
4.
5.
6.
Basics of Image processing
Medical Image modalities
Reconstruction
Registration
Segmentation
Enhancement
Mohammad Dawood
3
Medical Imaging, SS-2011
Image processing
Signal processing with an image as an input and an
image or a set of features as output.
Definitions
Image
Domain
In the discrete case
Mohammad Dawood
4
Medical Imaging, SS-2011
Classical methods of image processing include
Grayscale transformations
Color spaces
Filtering
Edge detection
Morphological operations
Mohammad Dawood
5
Medical Imaging, SS-2011
Grayscale transformations
The human eye can distinguish between different colors
with estimates ranging from 100,000 to 10 million!
Mohammad Dawood
6
Medical Imaging, SS-2011
Michelson contrast :
Weber contrast:
Mohammad Dawood
7
Medical Imaging, SS-2011
Grayscale Transforms
Mohammad Dawood
8
Medical Imaging, SS-2011
Grayscale transformations
Three of the most common
grayscale transforms are:
1. Linear
2. Logarithmic
3. Power law
Point operations
Mohammad Dawood
9
Medical Imaging, SS-2011
Linear color domain transform
X-Ray Mammogram
Mohammad Dawood
10
Medical Imaging, SS-2011
Power law
MRI of Spinal cord
Mohammad Dawood
11
Medical Imaging, SS-2011
Power law
CT of Head
Mohammad Dawood
12
Medical Imaging, SS-2011
Histogram
Histogram function :
Probability function:
Cumulative histogram:
Mohammad Dawood
13
Medical Imaging, SS-2011
Histogram Equalization
MRI of Spinal cord
Mohammad Dawood
14
Medical Imaging, SS-2011
Histogram equalization
Mammograms
Mohammad Dawood
15
Medical Imaging, SS-2011
Adaptive/Local Histogram Equalization
Mohammad Dawood
16
Medical Imaging, SS-2011
Local Histogram Equalization
Mohammad Dawood
17
Medical Imaging, SS-2011
Use of color spaces
Mohammad Dawood
18
Medical Imaging, SS-2011
Use of different color spaces
The continuous spectrum visible to human eyes
Mohammad Dawood
19
Medical Imaging, SS-2011
Use of different color spaces
RGB (Red, Green, Blue)
Mohammad Dawood
20
Medical Imaging, SS-2011
Use of different color spaces
RGB (Red Green Blue)
Cardiac PET
Mohammad Dawood
21
Medical Imaging, SS-2011
Use of different color spaces
HSV (Hue, Saturation, Value)
Mohammad Dawood
22
Medical Imaging, SS-2011
Use of different color spaces
HSV (Hue, Saturation, Value)
Cardiac PET
S=1, V=1
V=1
S=1
Mohammad Dawood
23
Medical Imaging, SS-2011
Using different spectrums
Cardiac PET
Mohammad Dawood
24
Medical Imaging, SS-2011
Fourier Transform
Euler’s formula:
Fourier transform:
Inverse Fourier transform:
Mohammad Dawood
25
Medical Imaging, SS-2011
Fourier Transform
Respiratory signal
Mohammad Dawood
26
Medical Imaging, SS-2011
Mohammad Dawood
27
Medical Imaging, SS-2011
Fourier Transform
Convolution theorm
Mohammad Dawood
28
Medical Imaging, SS-2011
Spatial filtering
Mohammad Dawood
29
Medical Imaging, SS-2011
Spatial connectivity
2D
- 4 connectivity
- 8 connectivity
3D
- 6 connectivity
- 18 connectivity
- 26 connectivity
Mohammad Dawood
30
Medical Imaging, SS-2011
Spatial filtering (local operators)
Filters are used in image processing for various purposes
e.g. noise reduction, edge detection, pattern recognition.
0
7
3
-2 3
-1 8
3
5
-6
1
1
1
4
0
3
7
4
1
1
1
0
1
-5 0
-3
1
1
1
7
1
4
-8
6
f
0
* 1/9
Applied only to red cell
h
7
3
-2 3
-1 3
3
5
-6
4
0
3
7
4
0
1
-5 0
-3
7
1
4
-8
6
f*
(0*1+7*1+3*1-1*1+8*1+3*1+4*1+0*1+3)*1/9 = 3
Mohammad Dawood
31
Medical Imaging, SS-2011
Noise reduction
Averaging filter
3
3
3
3
0
3
5
3
3
0
3
3
3
3
0
0
0
0
0
0
0
0
0
0
0
*
1
1
1
1
1
1
1
1
1
*1/9
=
Applied only to red cells
3
3
3
3
0
3
3.2 3
3
0
3
3
3
3
0
0
1
1
0.7 0
0
0
0
0
0
Cardiac PET, averaging with 5x5
Mohammad Dawood
32
Medical Imaging, SS-2011
Median filter
Median = Middle value of the set
Example
- given
- sort
S = {1, 5, 2, 0, -3, 8, 0}
S = {-3, 0, 0, 1, 2, 5, 8}
median(S)= 1
What happens if |s| is even?
- given S = {1, 5, 2, 0, -3, 8, 0, -5}
- sort
S = {-3, -5, 0, 0, 1, 2, 5, 8}
median(S)= 0.5
Mohammad Dawood
33
Medical Imaging, SS-2011
Noise reduction
Median filter
3
3
3
3
0
3
3
3
3
0
3
5
3
3
0
3
3
3
3
0
3
3
3
3
0
3
3
3
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Mohammad Dawood
* median filter
Applied only to red cells
=
34
Medical Imaging, SS-2011
Noise reduction
Gaussian filter
Gauss function is defined as:
Mohammad Dawood
35
Medical Imaging, SS-2011
Noise reduction
Comparison
Original
Mohammad Dawood
Averaging (5x5)
Median(5x5)
Gaussian (5x5)
36

similar documents