detection, visualization, and identification of lung abnormalities in chest spiral ct scans
DESCRIPTION
Detection, Visualization, and Identification of Lung Abnormalities in Chest Spiral CT Scans. 3D CT Image Data. Visualize Whole lung tissues Using VTK. 8 mm. Making stochastic Model using Gibbs Markov Random Field. Apply ICM using Genetic and EM algorithm. Removing Background. - PowerPoint PPT PresentationTRANSCRIPT
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Detection, Visualization, and Identification of Lung Abnormalities in Chest Spiral CT Scans
Abnormality
Detection System
Removing Background
Making stochastic Model using Gibbs
Markov Random Field
Apply ICM using Genetic and EM
algorithm
Visualize Whole lung tissues Using VTK
Visualize Abnormal Tissues Using VTK
3D CT Image Data
8 mm
Registration
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Medical ImagingTypes of medical Imaging
1. X-ray ImagingAdvantage Cheap Disadvantage It is just a projection of an object
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Example of X-ray Imaging
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Example of X-ray Imaging
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
2. computed tomography (CT) Advantage
1. better Geometry of the scanned subject
2. Using CT we can build 3-D model of the scanned subject
3. Give high contrast between bones and soft tissues
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Disadvantage
1. Ct has harmful effect due to radiation dose (X-ray)
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Example of CT
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
3. Magnetic Resonance Imaging (MRI)
Advantage
1. Give high contrast of soft tissues
Disadvantages
1. Does not preserve the geometry of the scanned subject if it is compared with CT
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Example of MRI
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
4. Ultrasound Imaging
Advantage
1. Real Time Imaging
2. No harmful effect
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Example of Ultrasound Imaging
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Abnormality Detection System
Removing Background
Making stochastic Model using Gibbs
Markov Random Field
Apply ICM using Genetic and EM
algorithm
Visualize Whole lung tissues Using VTK
Visualize Abnormal Tissues Using VTK
3D CT Image Data
8 mm
Registration
Automated Lung Abnormality Detection System
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
System Design1. Preprocessing Data
Such as you can filter your images in order to reduce the noise
1. LPF 2. HPF 3. BPF
3. Median filter 4. Gaussian Filter
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Image 3 x 3 pixel
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
1. Remove the background Starting from the edge of the image, neighboring pixels are compared. Pixels having the same gray levels are removed (I.e., belong to the same region), while those differing are kept.
Original image Image after removing background
Original Image 3x3 pixels
Image 3 x 3 pixels after
applying the algorithm
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
LungChest Background
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
How To estimate the Initial Mean for Lung and Chest?
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
CT Slice Contain Abnormal Tissues
Abnormal tissues
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Abnormal tissues
Slice_No. 32
Slice_No. 33
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Abnormality Detection Criteria
Each Ring Shape will take three ranks
1. Radial uniformity (R)
2. Position of the ring shape relative to the center of right or left lung edge (P)
3. Connectivity between different slices (C)
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Remove the Normal Tissues
Detecting
ring shape
Compute The Total Rank (R) for Each
ring shapeR> 2
Abnormal
Tissues
Normal
Tissues
yes
No
Abnormality System detection
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
a. Removing the normal tissues
In order to remove the normal tissues of the lung, we will compute the histogram for each slice and search for its peak, and then remove all pixels beneath this peak.
Before Removing normal Tissues
After Removing normal Tissues
Histogram of the CT slice
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
c. Ranking
1. NR, measures the uniformity distribution of the edges.
2. NC, measures the connectivity that the pixel (x, y) appears in the same location in different slices
3. NP, each pixel given a rank NP reflecting its position relative to the center of the right lung or the left lung.
Total Rank (N)= NR + NC + NP
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
4. Results
(a) Original slice from a spiral CT scan of a patient
(b) Slice after removing the background
(c) Desired tissues (e) The isolated lungs
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
(f) Bronchi, bronchioles and abnormal tissues
(g) Abnormal tissues detected by our algorithm
(h) Manual detection by expert doctor
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
Building 3-D modelWe use VTK tool to build 3-D model for the whole lung tissues and abnormal tissues, bronchi, and bronchioles
3-D model for the whole lung
tissues
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
This Figure shows the abnormal tissues in the 3-D
Computer Vision Image Processing Laboratory
www.cvip.uofl.edu
More Results